Maximum likelihood semiblind linear multiuser interference suppression for CDMA systems

Multiple access interference (MAI) and intersymbol interference (ISI) are the major limitations of third generation CDMA mobile communication systems. In this paper we propose a novel approach to linear interference suppression based on the maximum likelihood (ML) principle. The method is termed semiblind because it simultaneously exploits the transmission of training sequences and the probability density function of the unknown transmitted symbols. A block-iterative algorithm based on the expectation-maximization (EM) approach is proposed in order to obtain the filter coefficients. Computer simulations show that the resulting multiuser receiver exhibits a performance close to the linear minimum mean square error (MMSE) receiver.